REVIEW OF SCIENTIFIC INSTRUMENTS
VOLUME 71, NUMBER 11
NOVEMBER 2000
Picosecond infrared optical parametric amplifier for nonlinear
interface spectroscopy
D. Bodlaki and E. Borgueta)
Department of Chemistry and Surface Science Center, University of Pittsburgh, 219 Parkman Avenue,
Pittsburgh, Pennsylvania 15260
共Received 27 June 2000; accepted for publication 3 September 2000兲
A tunable, narrow bandwidth, high peak power picosecond infrared 共IR兲 laser system is described.
The pump source is a picosecond Ti:sapphire regenerative amplifier seeded by a picosecond
Ti:sapphire oscillator. The pump bandwidth and pulse duration are tunable producing 4–5 ps, 5–4
cm⫺1 pulses at 1 kHz. IR pulses are produced by optical parametric generation 共OPG兲 followed by
optical parametric amplification 共OPA兲. Tuning is possible over the entire 1050–3300 nm region of
the IR, with energies in excess of 15 ␮J over most of the range. The temporal and spectral
characteristics of the IR pulses are reviewed with a particular focus on the sources of bandwidth
broadening in the OPG/OPA. Bandwidth optimization of the IR output is discussed. A spectral
filtering scheme results in less than 15 cm⫺1 IR bandwidth, suitable for nonlinear optical
spectroscopic applications. © 2000 American Institute of Physics. 关S0034-6748共00兲02012-8兴
difference frequency generation in nonlinear crystals7 and
Raman shifting 8,9 have also been used to produce tunable IR
for SFG.
Free electron lasers offer a number of attractive features
for IR⫹VIS SFG, not the least of which are their broad
tunability in the IR 共1–100 ␮m兲 and high pulse energy 共⬎20
␮J兲.10,11 However, the requirement to travel to a dedicated
facility has limited their use to a few groups.
The generation of spectrally narrow pulses requires long
temporal pulses. Nanosecond OPO/OPA approaches have
been used successfully by a number of groups.12 However, if
temporal resolution and signal levels are not to be sacrificed,
shorter pulses are required. Pulses that are a few picoseconds
long with bandwidth ⬍20 cm⫺1 provide the spectral and
temporal resolution, as well as the necessary signal levels
appropriate for most interface science questions.
Transform limited Gaussian 3 ps pulse generation has
been achieved by the Laubereau group.13 This requires the
use of a Nd:yttrium–lithium–fluoride 共Nd:YLF兲 synchronously pumped optical parametric oscillator as the seed
source for the OPA.14 A common drawback of Nd:yttrium–
aluminum–garnet 共Nd:YAG兲 systems is their low repetition
rate, typically 10 Hz. This reduces the signal generated per
second, an important consideration when the photon counting statistics dominate the signal to noise ratio.
The development of femtosecond kHz Ti:sapphire regenerative amplifiers in the early 1990s has led to an explosion in tunable IR and VIS optical parametric sources.15–19
These are typically based on a broadband, continuum generation seeding stage.15,18 Continuum generation is a very
effective process in the 10⫺13 s regime, requiring ⬍1 ␮J and
producing little sample damage. Spectroscopic SFG spectra
have been obtained by mixing the broadband femtosecond
IR with a spectrally narrowed visible pulse for IR⫹VIS
SFG.20 Recently, the Shen group showed that interferometric
SFG with broadband fs IR pulses can yield high resolution
共⬍7 cm⫺1) interface spectra.21 A potential disadvantage of
I. INTRODUCTION
The investigation of fundamental chemical and physical
dynamical processes requires optical sources that optimize
spectral and temporal resolution. This is particularly important for nonlinear optical spectroscopies such as second harmonic generation 共SHG兲 and sum frequency generation
共SFG兲 that are highly interface specific.1–4 Spectroscopic
SHG and SFG require wavelength tunable sources providing
high peak powers. Narrow bandwidth ⌬˜␯ is required to resolve closely spaced spectral features. Short pulses ␶ are required to probe fast dynamics and to provide peak powers
for efficient SHG/SFG. The Fourier transform limit ⌬˜␯ ␶ determines the optimal achievable compromise. A 100 fs
Gaussian pulse, for example, has a 150 cm⫺1 transform limited bandwidth. This is an order of magnitude greater than
the bandwidth desirable for condensed phase vibrational
spectroscopy.
We report a 1 kHz, picosecond optical parametric amplifier 共OPA兲, pumped by a picosecond Ti:sapphire regenerative amplifier, providing tunable infrared 共IR兲 in the 1000–
3500 nm range with less than 15 cm⫺1 bandwidth. This
device has the necessary spectral and temporal resolution to
address many interface issues. The OPA design offers the
potential to tailor the source temporal and spectral characteristics for each experiment.
It is useful to review the different sources used for surface nonlinear spectroscopy. These fall into three broad categories, ns-long ps, short ps-fs, and free electron lasers. The
earliest surface IR⫹visible 共VIS兲 SFG experiments used a
fixed wavelength ns visible 共532 nm of Nd:YAG兲 and a tunable ns IR laser 共CO2). 5 Soon thereafter OPAs were employed for SFG.6 In various forms, OPAs have been the
workhorse for IR generation for SFG experiments, though
a兲
Author to whom correspondence should be addressed; electronic mail:
borguet⫹@pitt.edu
0034-6748/2000/71(11)/4050/7/$17.00
4050
© 2000 American Institute of Physics
Downloaded 06 Sep 2009 to 129.32.36.173. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp
Rev. Sci. Instrum., Vol. 71, No. 11, November 2000
Picosecond optical amplifier
approaches that employ broadband 共⬎50 cm⫺1) IR is that the
output can excite multiple distinct oscillators simultaneously.
However, this inconvenience can be turned into an advantage; ultrashort broadband IR pulses are ideal for quantum
beat SFG spectroscopy.22
The Richmond group at the University of Oregon recently took advantage of the availability of picosecond kHz
Ti:sapphire regenerative amplifiers to produce narrow bandwidth tunable IR.23,24 There are several notable differences
between the Oregon system and the one reported here. The
Oregon pump is a 2 ps, 17 cm⫺1 source 关⬎2 times transform
limit 共TL兲 of a Gaussian pulse兴, while the present pump provides a 4–5 ps pulse, 5–4 cm⫺1 共1.4⫻TL兲. As the pump
spectral width ultimately limits the bandwidth of the OPA
output, a 5 cm⫺1 pump offers the potential for spectrally
narrower IR pulse generation than a 17 cm⫺1 source. The
Oregon seed is provided by a continuum, whereas the OPA
discussed here used parametric fluorescence. Continuum
generation in the ps domain is unstable, easily damaging the
medium in which it is performed and requires ⬎100 ␮J,
more than two orders of magnitude the energy that is required in femtosecond OPAs. This consumes energy that
could otherwise be employed for experiments.
II. THEORETICAL CONSIDERATIONS
A number of nonlinear crystals are available for infrared
generation. We have chosen KTP because it possesses a high
damage threshold 共80 GW/cm2 reported for 2 ps, 1 kHz 800
nm pumping兲,23 high nonlinearity, and wide transparency
range 共0.35–4 ␮m兲. The performance of a crystal depends on
its effective nonlinearity, d eff . Although KTP is a biaxial
crystal, it can be treated as a uniaxial crystal because the
refractive index n z 共1.8297 at 1 ␮m兲 is larger than n x
共1.7377兲 and n y 共1.7453兲, and n x and n y differ slightly.25 The
effective nonlinear coefficient d eff can be approximated by
for type I phase matching
d eff⬇1/2共 d 15⫺d 24兲 sin 2 ␪ sin 2 ␾ ,
共1兲
for type II phase matching
d eff⬇ 共 d 24⫺d 15兲 sin 2 ␪ sin 2 ␾
⫺ 共 d 15 sin2 ␾ ⫹d 24 cos2 2 ␾ 兲 sin ␪ ,
共2兲
where ␪ is the angle the propagation direction makes with
the z axis and ␾ is the angle in the xy plane of the crystal.25
For KTP the nonlinear coefficients are d 15⬃2.04 pm/V, d 24
⬃3.92 pm/V at 880 nm, close to our pump wavelength, 800
nm.26 Equations 共1兲 and 共2兲 suggest that type II phase matching offers higher nonlinearity than type I phase matching.
We have considered three cases of type II phase matching in
KTP crystal: phase matching in xz, xy, and yz planes. Phase
matching in xz and yz planes provides the widest tuning
range 共1–4 ␮m兲. Type II phase matching in the xz plane has
the highest d eff .27
Two polarization schemes are possible when phase
matching in the xz plane. Ordinary idler and extraordinary
signal are generated for ␪ ⬍52°. The opposite combination
of signal and idler waves can be generated for ␪ ⬎52° 共Fig.
1兲 with a higher nonlinearity but a reduced tuning range.
4051
FIG. 1. Theoretical and experimental tuning curve 共A兲 and theoretical values of effective nonlinearity d eff 共B兲 for type II phase matching in the KTP
xz plane. On the phase matching curve 共A兲, the circles are measured signal
共e兲 wavelength and squares are measured idler 共o兲 wavelength. The o wave
denotes ordinary wave, e wave denotes extraordinary wave. The line is the
calculated phase matching angles.
Taking account of all these factors, type II phase matching in
the xz plane was selected, with ␪ ⬍52° so that an ordinary
pump generates an extraordinary signal and an ordinary
idler.
III. EXPERIMENT
A. Characterization tools
The Ti:sapphire oscillator, regenerative amplifier, and
OPA were characterized in terms of output power, wavelength, bandwidth, and pulse duration. The wavelength of
OPA was measured by SHG, in single crystal quartz, sent to
a monochromator 共Acton Research, SP-300i, 300
groove/mm 共gr/mm兲 grating, blazed at 1 ␮m, 11 nm/mm
dispersion兲, and detected by a liquid nitrogen-cooled charge
coupled device 共CCD兲 共Princeton Instrument CCD30-11,
26⫻26 ␮m pixel size, 0.3 nm/pixel resolution兲. The bandwidth 关full width at half maximum 共FWHM兲兴 of the oscillator and regenerative amplifier output was measured using the
1800 gr/mm grating 共optimized for 0.35–1.1 ␮m兲 in the
monochromator and CCD 共0.03 nm/pixel resolution兲. The
signal and idler bandwidths were measured using the 300
gr/mm monochromator grating 共resolution is 2 nm at 1 ␮m兲
and infrared InSb 共1–5 ␮m兲 or InGaAs 共0.8–1.8 ␮m, Thorlabs兲 detectors. Power was measured by a power meter
共Thorlabs兲. The oscillator and regenerative amplifier pulse
width were determined by autocorrelation. The signal and
idler pulse width was measured by crosscorrelation of the
infrared pulses with 800 nm pulses in a 2 mm thick BBO.
Autocorrelation and crosscorrelation signals were detected
Downloaded 06 Sep 2009 to 129.32.36.173. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp
4052
Rev. Sci. Instrum., Vol. 71, No. 11, November 2000
FIG. 2. Diagram of the Ti:sapphire regenerative amplifier and optical parametric amplifier. The Ti:sapphire oscillator pumped by an Ar ion laser provides 2.4–3.5 ps pulses with 10 nJ/pulse energy at 76 MHz repetition rate
tunable 740–860 nm. Regenerative amplifier pumped by a Nd:YLF laser
amplifies pulses to 0.7 mJ/pulse energy at 1 kHz repetition rate. The output
of the amplifier is sent to the optical parametric amplifier for frequency
conversion to the infrared.
by a Si photodiode 共Thorlabs兲, connected to a boxcar interfaced with a computer. Beam diameters were determined by
knife-edge measurement.28
B. Ti:sapphire laser system
The seed pulses for the Ti:sapphire system are supplied
by a picosecond Ti:sapphire oscillator 共Coherent Mira 900
F/P兲 pumped by all lines 共485, 514.5, . . . nm兲 of an Ar⫹ ion
laser 共Coherent Innova 310C兲 共Fig. 2兲. The Ti:sapphire oscillator provides a 2.4–3.5 ps, 76 MHz pulse train, with ⬎10
nJ/pulse energy, tunable in the 750–850 nm wavelength
range. While the output of the oscillator is adequate for many
nonlinear optical experiments, the pulse energy is too low for
significant population transfer to excited states and the repetition rate is too high for some systems to recover to their
initial state between successive pulses. What is needed is an
increase of pulse energy and increase of the time between
pulses. Both of these are achieved in the regenerative amplifier. A single temporally stretched seed pulse is selected by a
Pockels cell 共Medox兲 and further amplified in Ti:sapphire
regenerative amplifier 共Alpha-1000, BMI-Coherent兲 pumped
by the second harmonic 共527 nm兲 of a Nd:YLF 共BMICoherent兲. To avoid optical damage in the regenerative amplifier so called chirped pulse amplification is applied to reduce the high peak power densities that would otherwise by
present due to short 共1–4 ps兲 pulse durations.29 Pulses are
stretched before amplification to pulse durations ⬎100 ps.
Subsequent compression after amplification returns the pulse
duration to close to the initial preamplifiction values. Security systems monitor the pulse parameters 共repetition rate and
bandwidth兲 before the pulse enters the stretcher and aborts
the amplifier operation when the seed has the potential 共loss
of mode locking, . . . 兲 to damage the optics in the amplifier.
After compression the regenerative amplifier output is centered around 800 nm, providing 4–5 ps pulses with ⬎700 ␮J
energy and 1 kHz repetition rate.
The oscillator and regenerative amplifier pulses were
temporally and spectrally characterized. Figure 3共A兲 shows
the autocorrelation trace of the Ti:sapphire oscillator pulse. It
is well described by a Gaussian pulse shape with 3.5 ps
FWHM, corresponding to a 2.5 ps pulse duration. The time–
bandwidth product sets a limit to the smallest possible bandwidth that can be achieved for a given pulse duration. For a
Gaussian pulse the time–bandwidth product is 0.441, giving
D. Bodlaki and E. Borguet
FIG. 3. Autocorrelation 共A兲 and spectrum 共B兲 of the Ti:sapphire oscillator
共solid circles兲 and the regenerative amplifier 共open squares兲. In the pulse
width measurement 共A兲, the solid line is a Gaussian fit, FWHM is 3.5 ps for
the oscillator, and 5.8 ps for the amplifier. The corresponding pulse duration
is 2.7 ps for the oscillator and 4.1 ps for the amplifier. In the bandwidth
measurement 共B兲, the solid line is a guide to the eye. The oscillator bandwidth is 0.37 nm, and the amplifier bandwidth is 0.37 nm.
0.35 nm as the transform limited bandwidth at 800 nm. The
bandwidth of the oscillator was measured to be 0.37 nm 关Fig.
3共B兲兴 indicating that the oscillator output is 1.1 times transform limited. For comparison, the autocorrelation trace and
spectrum of the amplified pulse are also shown in Fig. 3. The
amplified pulse is also described well by a Gaussian with 5.8
ps FWHM, corresponding to a pulse duration of 4.1 ps. The
bandwidth was measured to be 0.37 nm. This indicates 1.6
times transform limited amplifier pulses. While the amplifier
pulse does not appear to have been perfectly recompressed,
stretching and compression do not strongly modify the spectral parameters of the pulse.
The pulse duration of the Ti:sapphire oscillator seed can
be varied in the 2.4–3.5 ps range. Bandwidth changes accordingly from 0.37 to 0.27 nm. The output of the regenerative amplifier follows the pulse width and bandwidth
changes of oscillator seed. The pulse width changes from 3.8
to 4.9 ps, with accompanying bandwidth changes from 0.37
to 0.27 nm.
C. Optical layout of OPA
The OPA design adopted is based on a first stage of
parametric generation followed by several stages of amplification. Figure 4 shows the OPA optical layout. The output of
the regenerative amplifier is split into two pump arms to
pump two OPA stages. The first stage consists of a single
KTP crystal 共KTP1兲 in a double pass geometry. The second
FIG. 4. Layout of the optical parametric amplifier: 共HWP兲 half-wave plate,
共PBS兲 polarizing beamsplitter, 共L兲 lens, 共M兲 silver mirrors, 共DM兲 dichroic
mirrors, 共KTP兲 nonlinear crystals, 共GR兲 grating, 共PR兲 prism.
Downloaded 06 Sep 2009 to 129.32.36.173. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp
Rev. Sci. Instrum., Vol. 71, No. 11, November 2000
stage consists of one or two crystals 共KTP2 and KTP3兲 employed in single pass configuration. All KTP crystals are
5⫻6⫻10 mm and cut at ␪ ⫽47.2°, ␾ ⫽0°. The combination
of a half wave plate 共HWP1兲 and polarizing beam splitter
enables the power in each arm to be varied.
The first stage pump is horizontally polarized. It passes
through a telescope that reduces the beam diameter by 10:1.
The pump beam is directed to KTP1 by a dichroic mirror
共DM1兲. The dichroic mirrors 共DM1–DM4兲 reflect 800 nm
(R 800⬎99.5%) and transmit 1–4.2 ␮m (T ave⬎95%). The
beam diameter of the pump at KTP1 is 600 ␮m 共340 mW
average power, peak power density 28 GW/cm2). In the first
pass, IR photons at signal and idler wavelengths are generated by optical parametric generation 共OPG兲 in KTP1. The
generated IR passes through DM2 which transmits IR and
reflects 800 nm. The signal wavelength is sent back to KTP1
for amplification after diffraction off the grating 共GR兲 共600
gr/mm, blazed at 1 ␮m or 1200 gr/mm grating blazed at 1
␮m兲. The 800 nm pump light is reflected back from a silver
mirror 共M4兲 and combined with signal seed at DM2. M4 is
placed on a translation stage so that the signal and pump
pulses can be temporally overlapped 共delay 1兲. In the second
pass in KTP1, amplification of signal wavelength and generation of idler wavelength occur. The pump beam diameter
on the second pass at KTP1 is 1 mm 共300 mW average
power, 9 GW/cm2). The signal beam diameter is 200 ␮m.
After the first stage, combined signal and idler energies as
high as 4 ␮J/pulse are achieved. The amplification factor in
the second pass was determined to be 105 .
About 260 mW of pump is directed to the second stage
of the OPA through a delay line 共delay 2兲. It is horizontally
polarized after the HWP2 and passes through a 7:1 reduction
ratio telescope. The 800 nm pump is combined with IR from
the first stage 共KTP1兲. The beam diameter of the pump is
580 ␮m 共23 GW/cm2) at KTP2. The diameter of the IR beam
is 260 ␮m at KTP2. Currently, both signal and idler are used
to seed the second stage of the OPA. In general, seeding with
only one wavelength creates a simpler, more controllable
situation. Seeding with the idler has been reported to result in
higher idler energies and qualitatively better beam profile.23
The second stage can accommodate one or two KTP
crystals. When two KTP crystal are used, they are placed in
such a way that when angle tuning the OPA they rotate in
opposite directions. In this way, beam displacement caused
by the rotation of one crystal can be compensated by an
equal amount of rotation of the other crystal. KTP3 also
provides additional amplification as shown in Fig. 5.
After passing through KTP2 and KTP3, the 800 nm light
is separated from the IR by a dichroic mirror 共DM4兲. Further
filtering of the OPA output depends on the application.
The wavelength tuning of the OPA is achieved by rotation of the KTP crystals and the GR. The crystals and grating
are placed on rotating stages which are driven by stepper
motors 共0.03 mrad/step兲. During wavelength scanning a
computer program 共LABVIEW兲 controls the rotation stages
and data collection.
Picosecond optical amplifier
4053
FIG. 5. Signal (␭⬍1600 nm兲 and idler (␭⬎1600 nm兲 power as a function
of wavelength. Open squares were measured with a two-crystal configuration, filled squares with a three-crystal configuration.
IV. RESULTS AND DISCUSSION
A. Tuning behavior
The wavelength range and output power over this wavelength range are primarily determined by the properties of
the nonlinear crystal. The phase matching angles are determined from the wavelength dependence of the refractive indices, given by the Sellmeier equations.25 The theoretical and
measured phase matching curve providing the wavelength
produced versus orientation of the crystal, are shown in Fig.
1. The measured data reproduces well the predicted phase
matching angles calculated using published refractive
indices.27
The signal and idler output power as a function of wavelength is shown in Fig. 5. During the measurement no intermediate alignment of the OPA was made. The initial configuration contained two KTP crystals. The tuning range was
limited to the 1.2–2.4 ␮m range, although KTP should enable the generation of wavelengths as long as 4 ␮m.23
Similar tuning behavior was reported by Carrion and
Girardeau-Montaut and was attributed to a decrease of effective nonlinearity d eff as idler wavelength increased.30 The
calculated dependence of d eff on wavelength is depicted in
Fig. 1. To estimate the effect of the wavelength dependence
of d eff , the amplification in a single pass was calculated considering a 10 mm long KTP and 25 GW/cm2 共power density
at KTP2兲 pump pulse. In the strong gain regime, the amplification A is exponential in gain ⌫ over the interaction length
L:
共3兲
A⫽1/4e 2L⌫ .
The gain is principally determined by the effective nonlinearity d eff when perfect phase matching (⌬k⫽0) is assumed:
⌫ 0⫽
冉
8 ␲ 2 I p˜␯ s˜␯ i
c ⑀ 0n in sn p
冊
1/2
d eff ,
共4兲
where I p is the pump peak power density, ˜␯ s and ˜␯ i are the
wave number of signal and idler, n p , n s , n i are the refractive indices of pump, signal, and idler, respectively, and d eff
is the effective nonlinear coefficient.27,31 Under these conditions, amplification varies by 6 orders of magnitude over the
entire tuning range of KTP 共1–4 ␮m兲 共Fig. 6兲. Equations 共3兲
and 共4兲 depict an ideal situation where losses due to im-
Downloaded 06 Sep 2009 to 129.32.36.173. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp
4054
Rev. Sci. Instrum., Vol. 71, No. 11, November 2000
D. Bodlaki and E. Borguet
FIG. 6. Estimated amplification for KTP type II phasematching in the xz
plane. Calculation is based on Eqs. 共3兲 and 共4兲 taking pump power 25
GW/cm2, crystal length 1 cm.
proper spatial and temporal overlap 共walkoff and group velocity mismatch effects兲 are not considered. The experimentally observed behavior 共Fig. 5兲 suggests that the OPA is
operating in the saturated regime. Such behavior has been
observed previously by others.27
Other possible factors that limit the tuning range were
considered. These include transmission of dichroic mirrors
共DM1–DM4兲, loss of 800 nm pump due to reflection from
the crystal surface, and efficiency of diffraction in first order
from the grating. These factors were carefully investigated
and optical elements were chosen so that these losses are
minimized. In addition, it is not expected that such factors
would show the strong wavelength dependence depicted in
Fig. 5.
Equation 共3兲 shows that the longer the interaction length
L, the more amplification is achieved. Our conclusion is that
the limited wavelength range is due to insufficient amplification. This suggests that in order to achieve wider wavelength tuning range, three or four KTP crystals would be
necessary. We added a third crystal to test this hypothesis.
The OPA efficiency improved and the tuning range expanded by more than 200 nm to longer idler wavelengths
共Fig. 5兲.
B. Bandwidth measurement
To resolve spectroscopic features a narrow-bandwidth
source is necessary. For instance, the vibrational spectra of
air/octadecyltrichlorosilane/glass interface shows modes
typically about 10 cm⫺1 wide and separated by 10–100 cm⫺1
from each other.32 Sources with bandwidth larger than 10
cm⫺1 would broaden features and might not resolve separate
modes. Even broader pulses, such as provided by fs lasers,
would excite several oscillators simultaneously.
The spectral bandwidth of the OPA output is determined
by several factors, including the gain bandwidth, the pump
divergence, and the divergence of the generated signal and
idler beams.31,33 When phase matching is not perfect (⌬k
⫽0), the gain is reduced. The FWHM of the gain curve
共gain versus ⌬k) is called the gain bandwidth (⌬k 1/2). It was
estimated to be about 46 cm⫺1 at idler wavelength ␭ i
⫽1920 nm using Eq. 共5兲:
FIG. 7. Bandwidth of pump and signal, idler wavelength: 共A兲 spectra of
pump. Bandwidth of pump is 5 cm⫺1, 共C兲 bandwidth of idler when there is
no grating in OPA. In this condition idler bandwidth is 68 cm⫺1. For comparison, 共B兲 and 共D兲 show signal and idler bandwidth with 1200 gr/mm
共blazed at 1 ␮m兲 grating in OPA. Signal and idler bandwidths are 14 and 17
cm⫺1, respectively.
1
⌬k 1/2⫽ 共共 2 ln 2 兲共 2⌫ 0 L 兲兲 1/2,
L
共5兲
where L is the interaction length and ⌫ 0 is defined by Eq.
共4兲.33 The pump beam divergence contributes to the bandwidth as follows. The finite divergence ␾ of the pump means
that the phase matching angle varies spatially across the
pump diameter. As a result, a range of signal-idler wavelengths is generated. The pump beam divergence was measured to be ⬃1 mrad, giving rise to 8 cm⫺1 bandwidth contribution. The generated signal beam also has a finite
divergence. This enables noncollinear phase matching, generating wavelengths different from those of collinear phase
matching. The signal divergence can be approximated by ␣
⫽d/L, where d is the pump beam diameter and L is the
crystal length.31 Considering a 580 ␮m pump beam diameter
共at KTP2兲 and L⫽1 cm crystal, a bandwidth contribution of
33 cm⫺1 is calculated.
Taking into account all these contributions 共gain bandwidth, pump divergence, and divergence of signal beam兲, the
overall bandwidth is determined to be ␦˜␯ ⬇( ␦˜␯ 21 ⫹ ␦˜␯ 22
⫹ ␦˜␯ 23 ) 1/2⫽(462 ⫹8 2 ⫹332 ) 1/2⫽62 cm⫺1 at idler wavelength
␭ i ⫽1920 nm. The idler bandwidth before any attempt to
narrow the bandwidth is 68 cm⫺1 共Fig. 7兲. This is consistent
with our estimates.
There are a number of ways to control the bandwidth.
Lowering the pump power reduces the gain bandwidth. A
proper choice and adjustment of the telescopes can reduce
the pump divergence and the pump beam diameter. Another
means to control the bandwidth is the use of a bandwidth
narrowing element, such as a grating, between the OPG and
amplification stages.
The infrared output generated in the first pass through
KTP1 was sent to a grating to narrow the seed bandwidth.
The signal wavelength was diffracted off the grating and
Downloaded 06 Sep 2009 to 129.32.36.173. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp
Rev. Sci. Instrum., Vol. 71, No. 11, November 2000
Picosecond optical amplifier
4055
while the signal pulse is five times transform limited for
1200 gr/mm grating.
D. Future directions
FIG. 8. Crosscorrelation of signal and idler pulses. Solid line is Gaussian fit.
FWHMs are 6.4 and 6.8 ps; corresponding pulse duration are 4.9 and 5.2 ps
for signal and idler pulses, respectively.
directed back to KTP1. The grating was placed in Littrow
geometry where the first order diffraction is sent back in the
direction of the incident beam. The Littrow geometry was
favored over Litmann–Metcalf geometry34 because of the
ease of alignment but also because grating performance was
optimized for this geometry. Gratings with groove density of
600 gr/mm 共blazed at 1 ␮m兲 and 1200 gr/mm 共blazed at 1
␮m兲 were tested. The bandwidth falling within the overlapping area of the pump and the diffracted signal at KTP1 can
be estimated from the grating equation
d
D
␦˜␯ ⬇˜␯ 2 a cos ␣ ,
共6兲
where a is the groove spacing, d is the beam diameter of the
IR beam, D is the distance between the grating and the KTP
crystal and ␣ is the angel of incidence on grating. At a signal
wavelength of 1370 nm, Eq. 共6兲 shows that a bandwidth of
9.1 cm⫺1 共600 gr/mm兲 or 2.8 cm⫺1 共1200 gr/mm兲 is achievable with these gratings. The pump bandwidth, 5 cm⫺1 共Fig.
7兲 limits the best achievable bandwidth in the latter case. The
measured signal bandwidth is reduced from 45 to 19.5 cm⫺1
when a 600 gr/mm grating is used in OPA 共not shown兲. The
1200 gr/mm grating further reduces the measured signal
bandwidth to 14 cm⫺1as shown in Fig. 8共B兲. The idler bandwidth with the same grating is 17 cm⫺1 关Fig. 7共D兲兴. This is
approximately three times broader than the pump bandwidth.
The signal/idler bandwidth is not limited by the grating or
the pump bandwidth. The preceding analysis suggests that
further improvement can be made by controlling the pump
beam diameter and divergence. This should be achievable
without sacrificing the power performance of OPA.
C. Pulse width measurement
The pulse duration of the output signal and idler were
determined. The signal and idler pulse widths were measured
by crosscorrelation of the signal and idler with 800 nm in a 2
mm BBO crystal. Figure 8 shows the crosscorrelation of the
signal and idler pulses. The Regen pulse width, assuming a
Gaussian pulse shape, is 4.1 ps 共Fig. 3兲. The signal and idler
pulse width were extracted by assuming convolution of two
2
⫽dt 2p ⫹dt 2i ). A crosscorrelation FWHM of
Gaussians (dt cc
6.4 ps 共6.6 ps兲 was obtained for the signal 共idler兲, resulting in
4.9 ps 共5.2 ps兲 pulsewidth. The signal and idler pulses are
about 1.3 times longer than the pump pulse.
Assuming Gaussian pulse shapes for the Regen and OPA
output, the Regen input pulse is 1.4 times transform limited,
The three-crystal configuration provides sufficient power
for spectroscopic and dynamic studies. However, intermediate alignment of the amplifier might improve efficiency in
ranges farther from the normal incidence position of the
crystals (␭ idler⬇1920 nm兲, where the OPA is optimized before tuning. As discussed, the bandwidth can be further narrowed by control of pump beam diameter and focusing.
The quality of the input beams is important in nonlinear
optical processes.24 Strong focusing of the pump beam in the
OPA crystals results in elliptical beam shapes and reduces
the efficiency of second harmonic generation. In the current
second stage configuration, the distance between the two
crystals 共KTP2 and KTP3兲 is fixed. An adjustable distance
between the crystals would give more control of focusing on
the individual crystals and consequently ore control over the
output beam quality. Beam diameter could be better matched
with other telescope reduction ratios.
In conclusion, an automated picosecond infrared optical
parametric amplifier tunable in the 1.1–3.3 ␮m range with
approximately 15 cm⫺1 bandwidth has been reported for the
purpose of combined spectroscopic and dynamic studies.
The OPA is based on three KTP crystals, all type II phase
matched in the xz plane. A grating is used to minimize the
output bandwidth to about 15 cm⫺1 making the OPA suitable
for spectroscopic studies. Without a grating, a bandwidth
⬎60 cm⫺1 was measured. Wavelength tuning of the OPA is
achieved by automated angle tuning of the nonlinear crystals
and the grating. The pulse width of the output was measured
to be close to the pump pulse width, indicating that the parametric process does not effect pulse width significantly in
this design.
ACKNOWLEDGMENTS
The authors would like to thank two REU students,
Andy Vidan and An Ngo, for their contributions. The REU
program is supported by NSF Grant No. PHY95-31628. This
work was supported by NSF Grant No. CHE-9734273.
Y. R. Shen, The Principles of Nonlinear Optics 共Wiley, New York, 1984兲.
R. M. Corn and D. A. Higgins, Chem. Rev. 94, 107 共1994兲.
3
K. B. Eisenthal, Chem. Rev. 96, 1343 共1996兲.
4
G. L. Richmond, Anal. Chem. 69, 536A 共1997兲.
5
X. D. Zhu, H. Suhr, and Y. R. Shen, Phys. Rev. B 35, 3047 共1987兲.
6
X. D. Hunt, P. Guyot-Sionnest, and Y. R. Shen, Chem. Phys. Lett. 133,
189 共1987兲.
7
D. Zhang, J. H. Gutow, K. B. Eisenthal, and T. F. Heinz, J. Chem. Phys.
98, 5099 共1993兲.
8
A. L. Harris, C. E. D. Chidsey, N. J. Levinos, and D. N. Loiacono, Chem.
Phys. Lett. 141, 350 共1987兲.
9
G. R. Bell, S. ManningBenson, and C. D. Bain, J. Phys. Chem. B 102, 218
共1998兲.
10
A. Peremans, A. Tadjeddine, P. Guyot-Sionnest, R. Prazeres, F. Glotin, D.
Jaroszynski, J. M. Berset, and J. M. Ortega, Nucl. Instrum. Methods Phys.
Res. A 341, 146 共1994兲.
11
M. Barmentlo, G. W. Thooft, E. R. Eliel, E. W. M. Vanderham, Q. H. F.
Vrehen, A. F. G. Vandermeer, and P. W. Vanamersfoort, Phys. Rev. A 50,
R14 共1994兲.
1
2
Downloaded 06 Sep 2009 to 129.32.36.173. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp
4056
Rev. Sci. Instrum., Vol. 71, No. 11, November 2000
R. A. Walker, B. L. Smiley, and G. L. Richmond, Spectroscopy 共Amsterdam兲 14, 18 共1999兲.
13
K. Wolfrum, J. Lobau, and A. Laubereau, Appl. Phys. A: Mater. Sci.
Process. 59, 605 共1994兲.
14
K. Wolfrum, H. Graener, and A. Laubereau, Chem. Phys. Lett. 213, 41
共1993兲.
15
V. V. Yakovlev, B. Kohler, and K. R. Wilson, Opt. Lett. 19, 2000 共1994兲.
16
P. E. Powers, R. J. Ellingson, W. S. Pelouch, and C. L. Tang, J. Opt. Soc.
Am. B 10, 2162 共1993兲.
17
V. Petrov and F. Noack, J. Opt. Soc. Am. B 12, 2214 共1995兲.
18
S. R. Greenfield and M. R. Wasielewski, Opt. Lett. 20, 1394 共1995兲.
19
B. Akhremitchev, C. F. Wang, and G. C. Walker, Rev. Sci. Instrum. 67,
3799 共1996兲.
20
L. T. Richter, T. P. Petralli Mallow, and J. C. Stephenson, Opt. Lett. 23,
1594 共1998兲.
21
J. A. McGuire, W. Beck, X. Wei, and Y. R. Shen, Opt. Lett. 24, 1877
共1999兲.
22
D. Star, T. Kikteva, and G. W. Leach, J. Chem. Phys. 111, 14 共1999兲.
23
D. E. Gragson, D. S. Alavi, and G. L. Richmond, Opt. Lett. 20, 1991
共1995兲.
12
D. Bodlaki and E. Borguet
24
D. E. Gragson, B. M. McCarty, and G. L. Richmond, J. Opt. Soc. Am. B
13, 2075 共1996兲.
25
R. W. Munn and C. N. Ironside, Principles and Applications of Nonlinear
Optical Materials 共Blackie Academic & Professional, London, 1993兲.
26
H. Vanherzeele and J. D. Bierlein, Opt. Lett. 17, 982 共1992兲.
27
J. D. Bierlein and H. Vanherzeele, J. Opt. Soc. Am. B 6, 622 共1989兲.
28
A. E. Siegman, in Solid State Lasers: New Developments and Applications, edited by M. Inguscio and R. Wallenstein 共Plenum, New York,
1993兲, p. 13.
29
P. Strickland and G. Mourov, Opt. Commun. 56, 219 共1985兲.
30
L. Carrion and J. P. Girardeau-Montaut, Opt. Commun. 152, 347 共1998兲.
31
A. Seilmeier, K. Spanner, A. Laubereau, and W. Kaiser, Opt. Commun.
24, 237 共1978兲.
32
P. Guyot-Sionnest, R. Superfine, J. H. Hunt, and Y. R. Shen, Chem. Phys.
Lett. 144, 1 共1988兲.
33
J. Y. Zhang, J. Y. Huang, Y. R. Shen, and C. Chen, J. Opt. Soc. Am. B 10,
1758 共1993兲.
34
M. G. Littman and H. J. Metcalf, Appl. Opt. 17, 2224 共1978兲.
Downloaded 06 Sep 2009 to 129.32.36.173. Redistribution subject to AIP license or copyright; see http://rsi.aip.org/rsi/copyright.jsp